You forgot from another forum that I'm putting a resistor on the USB gnd and ferrite choke.
But if you have a special PCI USB card like me - you don't need that either.
This has a battery on + and -
Last edited:
Hi
Yes again thermometer coding. But from the note it is first 4 MSB. N=4 (2**N)-1=15.
So AD1865 has first 15 MSB "DSD" and 18-4=14 LSB R2R. Almost half of the MSBs are from segmented part.
.
Since AD1862 from internal structure have 3 MSB N=3, (2**N)-1=7 encoded to Thermometer code.
AD1862 have frst 7 MSB "DSD", and 20-3=17 LSB are R2R.
.
For first 5 MSB binary - 31 MSB thermomether... To havy and complicated to implement
Imagine how long will be segment form all 16 bits encoded from binary to thermomether.
(2**16)-1=65535 bits word...
First 3 bits are "divided" into 7 parts of thermometer code - this does not touch the rest 17 bits ... anyway thermometer DACs are very good but almost impossible for more than 12-bits (very complicated) - that is where the R-2R technique can continue. AD1862 is 3 MSBs as thermometer + 17 LSBs as R-2RActually that is explanation that AD1862 is NOT pure R2R dac.
But segmented DAC or hybrid DAC.
Because first 3 bits are transformed as 7 bits of Thermometer code. And going to conversion same as DSD.
Other low significant bits are classic R2R dac...
So most of the sound comes grom generated first 7 bits
As other AD and PCM dacs...
This is interesting technique, but I am not sure if it is possible to create it "simple" from digital ICs. It looks like job for CPLD
The first MSB is very important and must be shifted in the correct place in DAC
XMOS is fast and very good but important is how the clocks are implemented. I like the implementation of this clock: http://pavouk.org/hw/audiosystem20/at32uc3a3256usbi2s.html ... (digital isolator should be omitted to achieve very low jitter).
I have been (very) slowly populating the small PSUs that @Traktorist3d shared a while back. Small dual rail PCB that are about 50x35mm at a guess just now so quite small. I'm not sure of the topology even if it is a shunt or not ...I think it's more a nazar type. Haven't tested yet but should perform well
@miro1360 Can I use electrolyte for PSU on 16V? (I need 5V on this time). I could't find nichicon 4700 and 6800 but instead Panasonic FC or Epcos.
Also I planning to solder np0 smd (GRM31C5C1H104JA01L) instead Wima on Dac board (1702) and instead Kemet (6pcs). This is for sample (I can get it back always).
Also I planning to solder np0 smd (GRM31C5C1H104JA01L) instead Wima on Dac board (1702) and instead Kemet (6pcs). This is for sample (I can get it back always).
Yes but no one said that other last significant bits are corrupted? Just want to point out the facts based on the data.First 3 bits are "divided" into 7 parts of thermometer code - this does not touch the rest 17 bits ... anyway thermometer DACs are very good but almost impossible for more than 12-bits (very complicated) - that is where the R-2R technique can continue. AD1862 is 3 MSBs as thermometer + 17 LSBs as R-2R
So the point is that AD1865 have more precise first 3 MSBs than AD1862 that have only 3 MSBs.
And there are NO R2R components like resistors. Current sources or some other active modules emulating R inside. R, C component are complicated to implement in IC...
this is science fiction to accomplish 12Bit Binary to thermometer code
it deserves (2**12)-1=4095 taps to go to R network and analog out "resistors" and very complex circuit.
For instanse we dont have any "chip" with 4095 pins and more if we want to get out with some ladder?
That is why manufacturers chose smaller number for thermometer segment.
.
True R2R ladder net, with pasive resistors, we can find just in discrete dac modules
.
For instance TDA1540 and TDA1541 are in BJT technologies, most other are witn MOSFET or mixed thech. Contemporary models are almost all with MOSFET tech. That is also can determine the sound?
.
Does not matter at the end everyone can choose by the listening what to use in reproduction
I use the oscillator board together with a gen 1 Xmos board in one of my dacs. With good clocks it provides obvious improvements upon the Xmos board, much as can be expected. The reclocking device remains the d-type on the Xmos board and the oscillator board provides merely a better regulator for the clocks and clock switching. If you have the current V3 Xmos, reclocking will still be provided by the fpga.
@sworder84
Raw DC voltage is important. For 16VDC filter capacitors, take 9VAC secondary transformers. For 5VDC at the output this is more than enough and the raw DC will not be over 14V at the first capacitor in the CRC. The AC voltage to the secondary is given for full load, if the transformer is less loaded that voltage rises above the rated. In any case, I recommend 25VAC filter capacitors for the first capacitor that is the most loaded. Check the spacing of the terminals on the capacitors to fit with the PCB.
The scheme published in post # 4,190 is wrong, the corrected scheme is attached. One transformer with two 12VAC secondary cannot be used with all positive regulators, but one transformer with 2x12VAC secondary and one with 2x9VAC secondary. With a combination of positive and negative regulators, only one transformer can be used. And LT1117T-5.0 TO220 were used instead of LP2950ACZ-05 TO92. I don't like those little ones in the TO92 case.
PS
Capacitors 4700uF 25V can be used for + -5V power supply part. I do not recommend 16V capacitors.
This is a temporary solution, but it works.
Attachments
Last edited:
Yes I had a note about this too earlier... A few posts before. It is more like Time simoultaneous format trigering conversion not like Rigt Justified.
And You dont have to delay the LR latch. Just o cut out all + statec prior to the event when LE going from 1 to 0. Only that ste triggers conversion.
And It should not be imeadetly after LSB. From LSB to LE should be 3-4 BCK cycles.
There are plenty of space to put the words inside LE. With a stopped clock.
I will post some circuits...
Correct Your I2S bus. Please note that BCK of I2S is always going from 1 to 0 always when LE changed state.
So invert RCK line and repeat sim?
It is most important to understand that finishing register loading and LE changing state from 1 to 0 is happening almost instantly. And that is the cae ONLY if Rigt Justified format used... Because of that only LE in this case has to be precise but other syncs are also important. And Jitter in time can have max afect.I did
... for generating new LRCK from a clean clock a ripple counter can be used (see the 74HC4040) https://pc.watch.impress.co.jp/docs/2004/1005/nidac_ah7.jpg
https://tinyurl.com/2zr7wcff
... better as delayed LRCK can be delay other data and the LRCK left as clean as possible (for the AD1862), ... this can be job for CPLD ... what do you think?
IF you left few bits of BCK when registers are already loaded, and LE stays in place, jitter could not be from that importance and DAc have much more time to "rest" and go to the conversion.
That is Whu Time Simultaneous format did this and Pacific Microsonic done the same - left blank sertain amount of BCKs prior to LE change state.
.
Only if we have 32 bit word we simply have to use all 32 bit for channel word and make it like pure Right Jusified format.
In all other cases we dont have to.
For 24 bit word we have 4 bits at each side of the word and 4 bits before each command to conversion.
What do you mean by "bits are corrupted"?Yes but no one said that other last significant bits are corrupted? Just want to point out the facts based on the data.
So the point is that AD1865 have more precise first 3 MSBs than AD1862 that have only 3 MSBs.
And there are NO R2R components like resistors. Current sources or some other active modules emulating R inside. R, C component are complicated to implement in IC...
this is science fiction to accomplish 12Bit Binary to thermometer code
it deserves (2**12)-1=4095 taps to go to R network and analog out "resistors" and very complex circuit.
For instanse we dont have any "chip" with 4095 pins and more if we want to get out with some ladder?
That is why manufacturers chose smaller number for thermometer segment.
.
True R2R ladder net, with pasive resistors, we can find just in discrete dac modules
.
For instance TDA1540 and TDA1541 are in BJT technologies, most other are witn MOSFET or mixed thech. Contemporary models are almost all with MOSFET tech. That is also can determine the sound?
.
Does not matter at the end everyone can choose by the listening what to use in reproduction
Thermometer from MSBs in AD1862/AD1865 is not for a higher precision, but rather for eliminating glitch. R-2R ladder is inside the AD1862/5 for the LSBs, is not?
Hard to say if the full R2R ladder is incorporated in any advanced R2R chips, even DAC8811 has the 3 MSBs segmented and the rest is R2R
Exactly, the listening test is the best way